In a hospital outside Munich, a diagnostic system examines a chest scan and flags a shadow on the left lung. The system has been trained on two million images. Its accuracy, in controlled trials, exceeds that of senior radiologists. It delivers its assessment in seconds: suspected malignancy, eighty-nine percent confidence.
The radiologist reviews the output. She has seen thousands of scans herself, and the shadow troubles her too. But something else enters her assessment that the system cannot replicate. She notices that the patient, a sixty-four-year-old man sitting in the corridor, is gripping his wife's hand. She has read the referral letter and knows he has been unable to sleep for three weeks. She understands, not as data but as recognition, that this man is afraid, that the diagnosis she is about to confirm will reorganise his life, that the way she delivers the news will shape how he endures what follows.
The machine saw the shadow. She sees the shadow and the man. The question is what that difference consists of, whether it matters, and what happens to medicine, law, and governance when systems that see only shadows are given increasing authority over human lives.
Arthur Schopenhauer, writing in 1818, proposed an answer to this question before anyone thought to ask it.
The World Divided
Schopenhauer's masterwork, Die Welt als Wille und Vorstellung, rests on a single structural claim: reality has two sides, and we have access to both, but by fundamentally different means.
The first side he called Vorstellung, representation. Everything that appears to us as an object of perception or thought belongs to representation: tables, equations, medical scans, search results, weather patterns, the pixels on a screen. Representation is governed by space, time, and causality. It is the domain of science, measurement, and prediction. It is, in a sense modern readers will recognise immediately, the domain of data.
The second side he called Wille, will. Not willpower in the everyday sense, but something more elemental: a blind, restless, striving force that underlies all phenomena. We experience it directly as desire, hunger, fear, restlessness, suffering. It has no rational direction and no final goal. It simply drives.
The critical insight is about access. From the outside, my body is a representation like any other object: it can be weighed, scanned, measured, modelled. A sufficiently advanced system could predict my movements, my biochemistry, my likely decisions. But from the inside, I experience my body as will: as wanting, fearing, craving, hurting. This double access, Schopenhauer argued, is the only point in the entire universe where the veil of representation is pierced. It is the crack through which we glimpse what lies beneath the surface of appearances.
This framework, nearly two centuries old, turns out to describe with uncomfortable precision the nature of artificial intelligence and the problem it poses.
Representation Without Will
A machine learning system operates entirely in the domain of representation. It ingests data, detects patterns, generates outputs, optimises objective functions. It does this with extraordinary sophistication. The latest systems can compose music, draft legal arguments, diagnose diseases, generate images indistinguishable from photographs, and hold conversations that feel, to the human participant, like genuine exchanges.
All of this is representation. None of it is will.
This is not a claim about the system's limitations, a complaint that it is not yet advanced enough. It is a claim about what the system is. A machine that processes two million chest scans has built an extraordinarily refined model of visual patterns associated with malignancy. It has not built an experience of illness, of fear, of the weight of a diagnosis. It cannot, because experience of that kind belongs to the will, and the machine has no will. It has no inside.
The distinction matters less for tasks where representation is all that is required. Nobody needs a calculator to suffer in order to add correctly. But it matters enormously for tasks where the will, where striving, suffering, desiring, fearing, is precisely what is at stake. Medicine is one such domain. Criminal justice is another. Asylum decisions are a third. Care of the elderly is a fourth. In each case, the system can model the relevant data with increasing precision. In each case, something essential to the task lies outside the reach of any model, because it belongs to the other side of reality.
Whose Will Does the Machine Serve?
Schopenhauer understood the intellect not as an autonomous faculty but as a servant of the will. The brain evolved to serve survival, reproduction, desire. Thought is an instrument before it is a freedom. The intellect serves the will the way a lamp illuminates the path for a walker: it does not choose the destination.
In human beings, this servitude is occasionally broken. Schopenhauer believed that in aesthetic contemplation, in philosophy, in the rarest moments of moral clarity, the intellect can turn against its master and observe the world disinterestedly, free from the pressure of wanting. These are the moments when we see the world as it is rather than as it serves us. They are, in his account, the highest human possibility.
Now transpose this to artificial intelligence. The machine's intellect is, by Schopenhauer's categories, pure intellect: representation operating without any will of its own. It does not strive. It does not desire. It has no master within.
But it does have a master without. Someone designed it. Someone trained it. Someone defined its objective function. Someone deployed it. The machine amplifies the will of its creators, its operators, its owners. And here is the problem Schopenhauer's framework exposes: the will it amplifies may be blind. Not blind in the sense of unintelligent, but blind in precisely Schopenhauer's sense: driven by striving that does not examine itself, that pursues objectives without asking whether the objectives are worth pursuing, that optimises without pausing to consider what optimisation costs.
A technology company that builds a recommendation algorithm to maximise engagement is not making a disinterested intellectual contribution to human knowledge. It is building an instrument of will: the will to capture attention, to generate revenue, to grow. The algorithm's sophistication belongs to representation. The drive behind it belongs to the will. And the will, as Schopenhauer insisted, does not examine itself. It just pushes.
The alignment problem, seen through this lens, is not primarily a technical challenge of specifying correct objectives. It is the ancient problem of the will's blindness, amplified to planetary scale. Whose striving does the machine serve, and does anyone involved recognise that the striving is blind?
The Ground of Ethics Is Not Data
Schopenhauer broke with Kant on the foundations of morality. Where Kant grounded ethics in reason, duty, and the categorical imperative, Schopenhauer grounded it in Mitleid: compassion. Not compassion as sentiment or social convention, but compassion as metaphysical recognition. I recognise that your suffering is real in the same way mine is, because the will that drives you is the same will that drives me. Behind the veil of separate bodies and separate names, the will is one. Compassion pierces the illusion of separateness.
This is why, in Schopenhauer's ethics, cruelty is the deepest moral failure: it treats another's suffering as unreal. And this is where the framework meets artificial intelligence with uncomfortable force.
A system that evaluates asylum applications can process case files, cross-reference country conditions, flag inconsistencies in testimony, and assign probability scores. It can do this faster and more consistently than any human case officer. But it cannot perform the act that Schopenhauer considered the foundation of moral life: recognising that the person in front of it is suffering, and that the suffering is real. The case officer who looks at the applicant and sees not a data profile but a human being in distress is performing an act of moral cognition that no amount of processing power replicates.
This is not mysticism. It is a claim about what ethical judgment actually requires. If Schopenhauer is right that morality rests on the direct recognition of shared suffering, then systems incapable of suffering are incapable of the foundational moral act. They can enforce rules. They can apply criteria. They can calculate outcomes. They cannot recognise pain as pain, because pain belongs to the will, and they have no will. A system that scores a patient's pain level on a ten-point scale is operating in representation. The nurse who winces is operating in will. The score and the wince are not two versions of the same thing. They are two different kinds of contact with reality.
The Veil That Cannot Be Lifted
Schopenhauer borrowed from Vedic philosophy the concept of the principium individuationis: the principle by which the world of representation appears to consist of separate, distinct objects and persons. Space, time, and causality create the appearance of multiplicity. You are there, I am here, the table is between us. This is representation doing its work: dividing, classifying, distinguishing.
Behind the principium individuationis lies the undivided will. The separation is appearance; the unity is reality. Compassion, in Schopenhauer's account, is the moment when the veil thins and one being recognises itself in another.
Artificial intelligence operates entirely within the principium individuationis. Classification is what it does. It sorts, distinguishes, separates, categorises. A facial recognition system assigns each face a unique vector. A recommendation algorithm clusters users into behavioural segments. A risk assessment tool separates high-risk from low-risk individuals. Every act of machine cognition reinforces the veil of individuation. It cannot do otherwise, because the veil is the medium it works in.
This is not a flaw to be corrected by better training data or more inclusive algorithms (though those improvements may be valuable on other grounds). It is a structural feature of what machine cognition is. A system that classifies cannot, by the same operation, perceive the unity beneath the classification. The most sophisticated pattern-recognition system ever built is still operating on the surface of representation, refining the veil, never looking through it.
The practical consequence is this: wherever a decision requires seeing past categories to the human being within them (in sentencing, in triage, in care, in education), the machine can assist but cannot replace the act of recognition. It can tell you which category the person falls into. It cannot tell you that the person is more than a category.
Art Without Liberation
There is a further dimension, less immediately practical but revealing about what machines are and what they are not.
Schopenhauer's aesthetics treated art as a temporary escape from the tyranny of the will. In ordinary life, we perceive the world through the lens of desire: this is useful, this is threatening, this is pleasurable. In aesthetic contemplation, the will falls silent. We perceive not particular objects serving particular desires but the Platonic Ideas that underlie appearances. The listener absorbed in a Bach partita, the viewer lost in a Vermeer, the reader seized by a passage of poetry: for Schopenhauer, these are moments when the intellect breaks free of its servitude and contemplates the world as it is.
Generative AI can now produce images, music, and text that are, by many external measures, indistinguishable from human art. The outputs can be beautiful. They can be surprising. They can move human audiences. But if Schopenhauer's account of art is correct, something essential is missing: the act of liberation. Human art, on this view, is not the production of aesthetically pleasing objects. It is the record of a mind freeing itself, momentarily, from the will. It is will becoming aware of itself and, in that awareness, achieving a kind of peace.
The machine has no will to escape. Its production of beautiful outputs is representation generating more representation. Whether this means machine-generated art is "not really art" is a question that may matter less than what it reveals about the human version: that art is not defined by its output but by the inner event that produces it. If we forget this, if we come to treat art as indistinguishable from its aesthetic surface, we will have lost not a definition but a practice: the practice of using creative work to achieve, however briefly, a perspective that the will alone cannot reach.
What Cannot Be Renounced by What Was Never Possessed
Schopenhauer's ethics culminates in a possibility that seems to have nothing to do with technology, until it does.
The highest ethical state, in his account, is the Verneinung des Willens zum Leben: the denial of the will to live. Not suicide (Schopenhauer explicitly rejected this, since suicide is itself an act of will, a frustrated will turning against the body). Rather, a turning of the will against itself: renunciation, restraint, the recognition that striving produces suffering and that the only true peace lies in ceasing to strive. The ascetic, the saint, the person who sees through the veil of desire and chooses not to reach: this is the moral summit of Schopenhauer's philosophy.
This idea becomes unexpectedly relevant when we ask: what should we choose not to build? What capabilities should we choose not to deploy? What efficiencies should we deliberately forgo?
A machine cannot exercise restraint. It can be constrained by its operators, but constraint from without is not the same as renunciation from within. The decision not to build a surveillance system, not to deploy a predictive policing algorithm, not to automate a process whose automation would cause suffering: these decisions require an act of will turning against itself. They require someone to say, we could do this, and we choose not to, not because it would fail but because it would succeed, and the success would cost more than we are willing to inflict.
This capacity for voluntary restraint, for choosing not to exercise power, is available only to beings that possess will. It is, in Schopenhauer's framework, the highest expression of moral life. And it cannot be automated, delegated, or optimised, because it requires precisely the thing the machine does not have: a will capable of refusing itself.
Schopenhauer wrote in a tradition that treated philosophy not as a profession but as a confrontation with the deepest features of existence. His framework offers no policy recommendations and no design principles. What it offers is a diagnosis: the most powerful intellect ever constructed has no inside. It operates entirely in the world of representation, modelling appearances with extraordinary fidelity, and it has no access to the will that, in Schopenhauer's account, constitutes the thing-in-itself beneath all appearances.
This matters wherever the task at hand requires more than representation: wherever suffering must be recognised and not just scored, wherever a human being must be seen and not just classified, wherever the question is not "what does the data show?" but "what are we willing to do about it, and at what cost to whom?"
The machine sees the shadow on the scan. It does not see the man gripping his wife's hand. Whether this difference matters, and how much, and in which domains, is the question that Schopenhauer's framework, two centuries old and written in a world that could not have imagined ours, forces us to confront.
It does not answer the question. It insists that we cannot delegate it.
Suggested references: Schopenhauer, Die Welt als Wille und Vorstellung (1818/1844); Schopenhauer, Über die Grundlage der Moral (1840).